1 Executive summary


Reductions in ED attendances for diabetes focused on lockdown 1. Attendance rates returned during ‘recovery’ phase with a small second reduction during lockdown 2 (Fig. 1.1). Considerable variation regionally but North East & Yorkshire and the Midlands regions were most impacted proportionally (Fig. 1.7). Coding issue lead to presentation of ED attendances for type 1 and type 2 diabetes separately, both excluding attendances for DKA.

Type 1 diabetes

  • Deprivation gradient seen in attendance frequency and covid-19 related reductions despite equally distributed prevalence of type 1 diabetes in the population. Proportionally, the most affluent quintile experienced the smallest reductions in attendance (Fig. 2.2.10).
  • Attendances were most frequent in under 40’s with the under 20’s having a significantly smaller reduction in attendance rates than other age ranges (Fig. 2.2.7).

Type 2 diabetes

  • Two distinct trends in cumulative and proportional attendance difference by region; sustained and growing attendance deficit in the Midlands, London and the North East & Yorkshire, while other regions cumulative reduction began to recover after lockdown 1 (Fig. 2.3.3).
  • Proportional attendance difference for the 40-59 years age group significantly lower than other age ranges due to the sharp return to above pre-pandemic rates of attendance, suggesting a backlog of activity was created during lockdown 1 which then presented during the recovery (see Fig. 2.3.6).
  • The Asian population experienced larger cumulative and proportional reductions in attendance rates than expected, mostly driven by reductions in ED attendances in Indian and Pakistani sub-groups (Fig. 2.3.12).

Inpatient DKA admissions

The seasonal and fluctuating nature of admission rates for DKA were not obviously impacted by the covid-19 pandemic (Fig. 2.4.1). Admissions were more frequent in type 1 diabetics at a ratio of approx. 3 in every 4 admissions for DKA, despite the higher prevalence of type 2 diabetes. The concerning combination of reduced ED attendances for less severe diabetes related emergencies and unchanged levels of or increases in inpatient DKA admissions were seen in:

  • Type 2 diabetics in London (Fig. 2.4.6),
  • 60-79 year old type 1 diabetics (Fig. 2.4.10) and
  • Pakistani and Caribbean ethnicities (Fig. 2.4.17).

The above may suggest hesitancy to attend ED at the earlier stages of an acute complication of diabetes is contributing to increased rates of more severe admissions for DKA. The inverse was seen in the most affluent 20% of the population, where lesser reduced levels of ED attendance in type 1 diabetes were seen with high reductions in DKA admission rates (Fig. 2.4.13).

2 Background

2.1 Diabetes care during a pandemic

The management of diabetes has long been spread across primary and secondary care providers and as such the effect of COVID-19 is less clear. It is conceivable that disrupted specialist input for type 1 and particularly high-risk type 2 patients has led to an increase in emergency attendance for hypo or hyperglycaemic events. Similarly, media and public perception of the stress upon the healthcare system may have reduced care seeking behaviour in individuals who did not feel safe attending A&E where they may have otherwise, potentially leading to worst outcomes when care was finally sought. Research from the Strategy Unit, Health Foundation and Nuffield Trust would suggest reduced care seeking would be disproportionately seen in certain ethnic minorities and cultural groups, thereby widening existing inequities in treatment and outcomes [1]

3 Methods

3.1 Data collection

3.1.1 Late diagnosis or acute complications of diabetes

Emergency attendances indicative of either late diagnosis or acute complications of diabetes were identified using SNOMED-CT diagnosis codes. Conditions included:

  • Hyperglycemia
  • Diabetic ketoacidosis (DKA)
  • Other acute metabolic complications of diabetes (hyperlipidaemia, hyperosmolarity and dyslipidaemia)
  • Coma due to diabetes (either direction, hyper/hypo)
  • Hypoglycemia
  • Diabetes related “findings” (“high blood sugar”, “increased glucose level”)

3.2 Project scope

The strategy unit proposes to quantify the changes in secondary care emergency admissions in diabetic patients according to similarly selected population groups.

3.3 Analysis plan

Plan:

  • Adjust SQL - diagnosis SNOMED-CT codes
  • Organise data in terms of:
    • Demographics (age, deprivation, ethnicity),
    • Geography,
    • Diabetes type,
    • Primary complaint and
    • Outcome (admission or not)
  • Present trends in above variables and explore further interactions

    Main unit of measure will be attendance change, comparing attendances between 2020 and 2019. 2019 data used as proxy for expected levels of demand in the absence of the pandemic.

4 COVID-19 impact diabetes activity

4.1 Emergency attendances


Fig. 1.1
Is the increasing trend a data quality issue or a valid pattern? Pre-covid activity rate will determine the extent to which covid has disrupted ‘normal’ treatment so validity is key.


Fig. 1.2
Consistent rates of planned and unplanned diabetes inpatient admissions suggest we would expect to see a flat trend in pre-covid ED attendances.


Fig. 1.3
By comparing yearly activity we can see that the 2019 trend serves as a useful counterfactual to compare 2020 activity against, as a proxy for what we would expect in the absence of covid-19. Data quality issues with the 2018 trend serves to lower the 2018-19 average considerably and artificially.


Fig. 1.4
Comparing the trend in ED attendances for diabetes related emergencies, we can see that there was a reduction in attendances in 2020 compared to 2019. The reduction was focused on the first national lockdown period and had returned to expected levels by the ‘recovery’ period.
We will assess whether avoided attendances were distributed evenly through the population subgroups.


Fig. 1.5
There is considerable variation in terms of pre-covid diabetes attendance volume and pandemic-induced reductions. The North East and Yorkshire and the Midlands regions have high pre-covid attendance volumes and considerable reductions during lockdown; both factors contributing to large cumulative reductions below.



Fig. 1.6
The cumulative difference in diabetes ED attendance for the South East is an outlier in that lockdown 1 saw a small reduction in ED demand with the increases during recovery period. This trend may suggest a backlog was created during the national lockdown that presented during the recovery phase.


Fig. 1.7
The highest levels of avoided or disrupted ED attendances were seen in the North East & Yorkshire and the Midlands, proportionally and in absolute terms. The North East & Yorkshire region saw 10% less demand for diabetes care in the ED than they would have expected, based on 2019 data.


4.2 Findings by condition

Fig. 2.1
Coding issues
60 snomed-ct diagnosis ID’s were used to search for diabetes-related ED attendances which identified 50-80 thousand attendances per year. Of the 60 diagnosis codes included in the data query, only the above 6 are present in the data. This may be expected given ED coding is in the gradual process of moving between systems (AEA to ECDS).
Only broad, high level snomed-ct ID’s appear to be in use, creating difficulties in identifying which type of diabetes the patient had when treated for DKA, for example. In this case, we have searched for secondary diagnosis codes which indicate the type of diabetes however almost 80% of records only have a primary snomed-ct code input. We attempted to use the record’s ‘Chief complaint’ field to identify diabetes type where not explicitly stated, however this reduced all diabetes-related attendances to either hypo- or hyper-glycaemia. Finally, we took a similar approach using the accompanying AEA diagnosis field, however again this did not offer enough granularity and classified diabetes-related records as either:

  • 301: Diabetes and other endocrinological conditions - Diabetic
  • 302: Diabetes and other endocrinological conditions - Other non-diabetic


As such, given the assumption that almost all DKA attendances would result in an admission, we will focus on inpatient data with regard to DKA where we can more accurately split data by diabetes type. We will therefore present trends in ED attendance for type 1 and Type 2 diabetes not including DKA.

4.2.1 Type 1 Diabetes

4.2.1.2 Age

Fig. 2.2.5
The frequency in which type 1 diabetes patients attend ED seems to reduce with increasing age. The most high volume age ranges are the under 40’s. Admission rates in those ages 40-99 however have remained consistent throughout the national lockdown periods.


Fig. 2.2.6
The cumulative difference is determined by the pre-covid attendance volume and the magnitude of the disruptive covid-19 impact. As such, the greatest cumulative reduction in ED attendance is seen in 20-39 year old’s, for which rates continued to drop during the recovery period and the second lockdown.



Fig. 2.2.7
Covid-19 had the least disruptive impact on ED attendances for the under 20 age range; this is due to the steep recovery after the first national lockdown, reducing the overall attendance deficit for this group.


4.2.1.3 Deprivation

Fig. 2.2.8
Given that the prevalence of type 1 diabetes is broadly similar across deprivation quintiles, we see inequalities here in the frequency at which more deprived populations require emergency treatment for type 1 diabetes.

https://jech.bmj.com/content/54/3/173
https://pubmed.ncbi.nlm.nih.gov/10975218/


Fig. 2.2.9
Increased demand for care in the more deprived groups translates to a increased cumulative reductions in ED care where we see more deprived populations avoiding, not accessing or not needing emergency care.


Fig. 2.2.10
Proportionally we see similar impacts across all of the deprivation quintiles other than the most affluent 20% of the population where we see less of a reduction in ED demand.


4.2.1.4 Ethnicity

Fig. 2.2.11
High levels of variation/fluctuation seen in non-white ethnicities due to small numbers of attendances - demonstrated by broad grey confidence intervals, as such conclusions are hard to draw from changes in trends.


Fig. 2.2.12
Reductions in ED attendances for type 1 diabetes occurred in each ethnic group. Differences in absolute cumulative reductions are driven by population demographics which much higher numbers of patients of white ethnicity.



Fig. 2.2.13
All ethnic sub-groups that demonstrate a proportional increase have such broad confidence intervals that limit our ability to draw conclusions.


4.2.2 Type 2 Diabetes

4.2.2.2 Age

Fig. 2.3.5
Given the increasing prevalence of type 2 diabetes, it is not surprising that the elderly age groups attend ED for diabetes related emergencies more frequently. In all age ranges, the pre-pandemic attendance volumes were achieved during the recovery period. The 80-99 year old’s appear to have experienced the slowest recovery, perhaps linked to the dangers of attending hospital for such high risk groups.


Fig. 2.3.6
A gradual and considerable cumulative reduction is seen in the oldest age ranges (60-99 years) while the 40-59 year old’s demonstrate a markedly different trend where attendances surpassed the previous year’s volume to reduce the number of avoided attendances - this perhaps indicative of a backlog built up during the initial lockdown.


Fig. 2.3.7
Proportional variability is seen between age ranges; 40-59 year old’s showing minimal overall disruption to ED care while 20% of attendances in the 80+ population were avoided or disrupted by the pandemic.


4.2.2.3 Deprivation

Fig. 2.3.8
A deprivation gradient is present in type 2 diabetes attendances to ED however it is unclear whether is is simply a function of the greater prevalence of type 2 diabetes in more deprived populations or indicative of further inequality in patient outcomes as well as disease prevalence.


Fig. 2.3.9
All deprivation quintiles experienced reduced demand for care during the first lockdown and plateaued during recovery; however, during the second lockdown we see further reductions in care in the most deprived 40% and increases in attendance for the lesser deprived quintiles 3 and 4.


Fig. 2.3.10
When considering proportional changes, we don’t see a statistically coherent deprivation gradient in terms of type 2 diabetes attendance.


4.2.2.4 Ethnicity

Fig. 2.3.11


Fig. 2.3.12
Though Black and Asian ethnic groups represent similar numbers of ED attendances for type 2 diabetes, the cumulative reduction is much higher in Asians. For the Black population, the reduction plateaued during the recovery phase and stayed level while Asian’s continued to not present at ED at the expected rate through recovery and the second lockdown phases.


Fig. 2.3.13
Significant reductions in Pakistani and Indian populations may have driven the cumulative trend seen above.


4.2.3 DKA - Inpatient admissions


Fig. 2.4.1
Inpatient admission rates for DKA had returned to pre-pandemic levels by the end of the recovery period. The behavioural drivers of DKA seem to have persisted around the Christmas and New Year’s period of 2020, despite the second national lockdown being in place. Seasonality is also seen in ED data but quality/completeness are issues prior to 2019.


Fig. 2.4.2
Though type 1 diabetics represent a much smaller population, they are responsible for more than twice the DKA inpatient admissions. DKA admissions in type 2 diabetics appear less impacted by the pandemic than those in type 1 diabetics.


4.2.3.1 Region


Fig. 2.4.3
Trends in DKA admission by diabetes type are similar and consistent with national trends.


Fig. 2.4.4
The cumulative difference in DKA admission varies by region, however all have experienced reductions during the pandemic. The East of England saw the sharpest decline in admissions yet the trend suggests there may have been a delayed effect as the curve is approx 5 weeks behind that of other regions.


Fig. 2.4.5
DKA admissions were least reduced in London and the South East with around only 2.5% less DKA admissions than previously seen in 2019. The East of England had the largest proportional reduction in inpatient DKA admissions which was significantly larger than any other region.


Fig. 2.4.6
When comparing change in ED attendances for diabetes (excluding DKA) and inpatient admissions for DKA, we would expect to see regions in either the bottom left or top right quadrants.
We hypothesize the following placements would suggest:

  • Bottom left: reductions in demand in both types of presentation
  • Top right: Smaller reductions or increases in demand in both types of presentation
  • Top left: Less disruption in ED attendance with reductions in DKA suggesting earlier presentation to ED and less severe complications
  • Bottom left: Reductions in ED presentation and no reduction or even increases in DKA admissions suggest delayed presentation and more severe outcomes.


If we accept these assumptions, we see type 2 diabetes in London experiencing delays in presenting to ED with less severe complications and no reduction in more sever DKA admissions.




4.2.3.2 Age


Fig. 2.4.7
DKA is distributed differently between age ranges when considering type 1 and type 2 diabetes; in type 1 diabetes, the younger age ranges are more frequently admitted for DKA while in type 2 diabetics, the middle age ranges represent the most at-risk group. It could be hypothesized that lifestyle and behvioural factors are drivers of increased rates in the 20-39 year old type 1 diabetes and familial or social care may reduce rates in 80-99 year old type 2 diabetics.


Fig. 2.4.8
The youngest age ranges (0-40 years) experienced a largest cumulative reduction in DKA inpatient admissions; where the under 20 admission rate reduced only during lock-downs that of the 20-39 year old’s continued to drop during the recovery period as pre-covid rates were not returned to post-lockdwn 1.


Fig. 2.4.9
Reductions in under 40’s are mostly driven by trends in DKA admissions in type 1 diabetics, while increased DKA admission rates in 40-79 years old’s are mostly a result of increased hospitalisations in type 2 diabetics.


Fig. 2.4.10
In type 1 diabetes, we see increases in DKA admissions in 40-59 and 60-79 year old’s. According to the above assumptions, we might conclude that increases in 60-79 year old’s with DKA were a function of reductions in less severe ED attendances.


4.2.3.3 Deprivation


Fig. 2.4.11
While inequalities in DKA admission are seen to exist between deprivation quintiles, they are lesser in Type 2 diabetes. In both cases, patients from the most deprived 20% of the population are responsible for the most DKA admissions.


Fig. 2.4.12
The impact of the first lockdown was most disruptive to inpatient admissions for IMD quintiles 1 & 2. The rate a which admissions reduced during the second lockdown was also much higher for the more deprived groups.


Fig. 2.4.13
While in absolute terms the more deprived groups saw greater reductions in DKA admissions, IMD quintile 5 experienced the largest proportional reduction however this is not statistically significant.


Fig. 2.4.14
In the most affluent 20% of type 1 diabetics, we see a significantly less disrupted rate of ED attendances along with a more reduced rate of DKA admissions comparatively. Changes in demand appear to be in line with the above assumptions in tpye 2 diabetes.



4.2.3.4 Ethnicity


Fig. 2.4.15
The seasonal nature of DKA admissions is most apparent in White groups and is consistent across diabetes types. Similarly, in both diabetes types, the rate of admissions in patients of “Unknown” ethnicity increases during the pandemic, perpahs suggesting compromised recording during times of high stress.


Fig. 2.4.16
As the magnitude of cumulative changes to DKA admission rates between ethinicities varies, so to does the direction; White groups saw reductions in rates limited to lockdown periods, while Black ethnicities experienced increased DKA incidence during lockdown periods with plateaus during the recovery period.


Fig. 2.4.17
All three sub-groups of the Black ethnicity were seen to demonstrate increased DKA admissions during the pandemic, with members of the Caribbean community admitted at a rate 25% higher than the previous year.


Fig. 2.4.18
In type 1 and type 2 diabetics, patients from Pakistani and Caribbean ethnic groups appear to be presenting to ED less regularly than would be expected for said groups but were experiencing much higher rates of DKA admissions (approx 10% and 30% increased respectively).


4.3 Attendances by severity


Fig. 3.1
By considering the duration of a patient’s ED atttendance and whether or not that attendance resulted in an inpatient admission, we can deduce a measure of attendance severity. We can see that the pandemic impacted the rate of attendances that were treated in the ED and discharged without inpatient admission different than other severity flags; at the peak of the first national lockdown, ED’s were more frequently treating patients and discharging from ED in under 4 hours, presumably to avoid further inpatient capacity stress and reduce covid risk to patients.


4.3.1 Type 1 Diabetes

Fig. 3.2.1
Fig 3.2.2
The first lockdown had similar impacts on the severity of type 1 diabetes attendances however the rate of return to pre-pandemic levels is lower in ED attendances that last between 4 and 12 hours and result in discharge from ED.


4.3.2 Type 2 Diabetes

Fig. 3.3.1
Fig. 3.3.2
Fig. 3.3.3
The first national lockdown saw a large increase in the rate of type 2 diabetic ED attendances being treated in ED in under 4 hours and discharged without requiring inpatient admission. This is a clear example of organisational behaviour change in the face of the pandemic to avoid admissions where possible.


4.3.3 DKA


Fig. 3.4.1
Length of stay is seen to generally be less frequent but of longer duration in DKA admissions for type 2 diabetics than for type 1 diabetics. One explanation may be that DKA in type 2 patients is more likely to occur in older individuals who are therefore more likely to have more significant comorbidities which would increase the wider care requirements. Interestingly the length of stay for type 2 diabetics with DKA was more impacted by the covid-19 pandemic as illustrated by reducing average LoS during the first lockdown; a time when hospitals were motivated to create capacity for incoming covid-19 patients, this trend isn’t seen in type 1 diabetics with DKA.



Fig. 3.4.2
In support of the above explanation, type 2 diabetics with DKA are seen to require critical care intervention more frequently than type 1 diabetics. Similarly, critical care usage in type 2 was more impacted or purposefully limited during the first lockdown.


5 Notes and references

Notes:

COVID-19 lockdown dates: https://www.instituteforgovernment.org.uk/sites/default/files/timeline-lockdown-web.pdf

References:

[1] The Strategy Unit., 2021. Strategy Unit analysis published showing changes in use of emergency departments under lockdown. Available at: https://www.strategyunitwm.nhs.uk/news/strategy-unit-analysis-published-showing-changes-use-emergency-departments-under-lockdown (Accessed 16th March 2021)